Net traffic ratings debated

SAN FRANCISCO--With a common belief that only one or two of them are likely to survive an imminent shakeout, four companies providing Web usage ratings took to the stage here last night in what sponsor Infoseek called the first "great Web ratings debate."

Like Nielsen Media's television ratings, which inform the public how many Americans tune into various broadcasts--and therefore determine how much ad time is worth to advertisers--Web ratings companies attempt to gauge how many and what types of people are looking at particular Web sites.

Nielsen's Web unit was on the panel, along with Media Metrix, RelevantKnowledge, and NetRatings. "You make or break the Web sites," observed panel moderator and newsletter Iconocast editor Michael Tchong. "Does it make you feel good to have that kind of clout?"

Given the power the firms have, the race to stay in the game until the market settles on one or two providers is worth a lot.

But although the Net ratings firms may feel good about their degree of influence, a more apparent attitude these days seems to be defensiveness over their methodologies.

The business of rating Web sites' traffic involves vexing problems, and differences in technique have yielded wildly varying results among the competing ratings firms. At last night's debate, company representatives sparred over how to collect a sample of Web surfers and how to study it.

The most common technique for recruiting subjects is to phone them by random-digit dialing, persuade them to load monitoring software on their computers, and then convince them to continue their participation in the study. Media Metrix, RelevantKnowledge, and Nielsen Web employ this method. Media Metrix also targets users by mail.

NetRatings drew the scorn of its competitors for its method of recruiting surfers on Web sites.

"A technique that uses a medium to measure that same medium is just not valid," said Manish Bhatia, Nielsen Web's vice president of interactive services.

Bhatia said his company had experimented with gleaning sample users from the Web rather than over the telephone, and despite the increased size of the sample, results were "way, way, way out of whack." Web-recruited surfers tended to be heavier users, disproportionately male, and with much more experience than the typical Web user.

"That number was rubbish," Bhatia concluded.

NetRatings vice president of marketing Tim Meadows defended the process of using the Web to gather sample users, saying his company "employed several steps in the recruiting process to make sure [it was] reaching a random sample."

One differentiating factor among the four groups is the coverage of people who access the Web from work, rather than home. Media Metrix and RelevantKnowledge currently are building business samples, and NetRatings and Nielsen Web have yet to do so.

The problem for the ratings companies is getting workers to cooperate with their studies. According to an internal Nielsen Web poll, only 9 percent of those contacted said they would install metering software on their workplace computer. That compared to 70 percent who said they would install such software at home.

And that's not the only obstacle to measuring the workplace Web surfer. Others include the hesitancy of information systems managers to install such software on employees' computers and the public relations nightmare of having employees' surfing habits exposed.

In one particularly embarrassing case, Penthouse magazine declared itself the most visited site on the Web after a Nielsen audit, thanks to frequent visits from workers at firms including IBM and Apple Computer.

One might think that a medium with cookies and server logs would simplify the task of tracking who goes where and how often. But it turns out that Internet technology has made things more difficult for ratings companies.

Server logs turn out to be misleading on a number of fronts. To conserve bandwidth, many companies cache Web sites that their employees visit frequently. That caching skews the information collected in server logs.

Another problem with server logs is that they don't measure demographic information that is extremely valuable to advertisers, such as their age, sex, and buying habits.

But computer technology has made things easier in some respects, said Nielson's Bhatia. In contrast to television ratings, which require a hardware installation at the subject's site, Web rating software can be downloaded easily.

One thing all the firms agree on is that in a few years' time, a similar panel will feature less than four players.

"I can see a future with two companies using different models or measuring different audiences," Bhatia told NEWS.COM. "But four? I doubt it. That creates confusion, and the market doesn't like confusion."